Dear managers, thanks to Casper.Dik@Sun.COM francisco roque <frisco@blackant.net> Tim Bradshaw <tfb@tfeb.org> "Matthew Stier" <Matthew.Stier@us.fujitsu.com> Markus Mayer <mymaillists@gmx.at> who suggested to use the following tools to see what is going on. vmstat -p 10 memory page executable anonymous filesystem swap free re mf fr de sr epi epo epf api apo apf fpi fpo fpf 18772512 4806232 774 1905 24 0 0 828 0 0 0 0 0 171 28 24 18861096 4880280 4072 706 13 0 0 35 0 0 8 0 0 540 14 13 ... iostat tty md0 md1 md3 md10 cpu tin tout kps tps serv kps tps serv kps tps serv kps tps serv us sy wt id 0 73 1673 117 27 9 0 39 21 1 34 1221 80 32 12 3 0 85 mpstat CPU minf mjf xcal intr ithr csw icsw migr smtx srw syscl usr sys wt idl 0 105 1 90 74 17 1431 57 299 73 2 2613 12 3 0 85 2 100 0 189 918 395 1434 44 304 116 2 2534 15 3 0 82 16 103 1 103 81 21 1467 61 294 85 2 2728 11 3 0 86 18 99 3 112 146 83 1663 66 333 86 2 3040 10 4 0 85 pmap -S 22659 ... -------- ------- ------- total Kb 3284520 3251312 An oracle tool (em) currently - I am still running a test with a process currently using 3.2 GB of RAM - gives me Memory utilization: 72 % "prstat -s size" make me thinking that only 6 GB or so of the 16 GB installed in the machine are used: PID USERNAME SIZE RSS STATE PRI NICE TIME CPU PROCESS/NLWP 22659 ahoesch 3208M 3194M cpu18 0 4 1:39:39 25% SmartClient/1 5950 oracle 1151M 1082M sleep 59 0 0:00:15 0.0% oracle/13 6793 oracle 1137M 1073M sleep 59 0 0:00:00 0.0% oracle/1 ... 6822 oracle 1137M 1077M sleep 59 0 0:00:00 0.0% oracle/1 5944 oracle 1137M 1073M sleep 59 0 0:00:03 0.0% oracle/1 27988 nbraun 465M 315M sleep 59 0 0:51:02 0.0% soffice.bin/6 ... I was told that the memory used by oracle is shared so that it actually uses 1150 MB only once. So this would be 3.2 GB for my test process 1 GB for oracle 5 GB estimated for all the other processes ===================== 9 GB Well, that's actually about 72 %. System performance is fine right now. It seems I can't reproduce the problem with my test process. Today a user with a 3.4 GB process bogged down the system completely. I will need to wait for the next real occurance and then recheck all the values. What numbers in the above stats should I look for first? Looking at "prstat -s size" did not get me anywhere. Looking at the output made me always think that there should be enough memory left. Is anything fine until the second column of "vmstat -p 10" output reaches zero? mpstat does not seem to help either, it shows 85% idle all the time. What values in iostat output should I have an eye on when the problem reoccurs? Thanks a lot, Andreas > a user process utilizing 2GB - 4GB of memory on a SF 490 (2 x SPARC IV, > 16 GB) causes serious performance problems for all users. It looks like > much disk i/o is taking place. The performance problem goes away 1-2 > minutes after killing the user process that consumed the 2GB - 4GB of > RAM. Since we have 16GB in the machine this should have no effect on > other users IMHO. No swapping should be performed since there is more > than enough physical memory to hold all process data in RAM. > > What could be the problem? I heard that Solaris does not like greater > portion of RAM being occupied by a single process. Is that correct? > Why? > > Thanks a lot, > > Andreas > _______________________________________________ > sunmanagers mailing list > sunmanagers@sunmanagers.org > http://www.sunmanagers.org/mailman/listinfo/sunmanagers _______________________________________________ sunmanagers mailing list sunmanagers@sunmanagers.org http://www.sunmanagers.org/mailman/listinfo/sunmanagersReceived on Fri Feb 1 13:11:37 2008
This archive was generated by hypermail 2.1.8 : Thu Mar 03 2016 - 06:44:10 EST